Block Matching and 3D Filtering (BM3D) for Preprocessing of CT scans of Covid-19 Lung Images

Abstract

SARS-CoV2 or the Corona Virus 2 is the cause of the global Corona Virus illness 2019 epidemic, also known as the COVID-19 pandemic (SARS-CoV-2). Wavelets are mostly used for denoising a two-dimensional signal for images mostly and so we have adopted a Discrete Wavelet Transform for CT image of covid and healthy lung images. Block Matching and 3D filtering can give better performance for pre-processing. Evaluation parameters such as SSIM, CC and NCC are considered for DWT and BM3D methods for Covid and Non-Covid models of imageries and evaluated. The image denoising using DWT resulted in an SSIM value of 0.564 for covid images and 0.6935  for healthy lung images, NCC of 0.997 and 0.998 for covid and healthy lung images respectively and CC of 0.9794 and 0.99234 for covid and non-covid images respectively. Image denoising using BM3D the SSIM values are computed as 0.919    and 0.926   for covid and healthy lung images respectively, NCC of 0.9996 and 0.999689 for covid and healthy lung images respectively and CC of 0.99286 and 0.9967 for covid and healthy lung images respectively. It has been observed that the BM3D method provides a good performance compared to the DWT Technique for both covid and healthy lung images considering the performance metrics.